Is post-testing after two-way ANOVA effective when the factor interaction is significant?

Prism only performs one kind of post test--if you select two columns, Prism compares the columns for each row, without regard to whether there is row/column interaction.

One could argue that this kind of post test only makes sense when there is significant interaction, that is, when the differences between columns is not the same for each row. When there is no significant interaction, Prism still does the same post test, but it may be that you should only look at the overall row and column effects.

We've recently added a web calculator that can do other post tests between pairs of cells in your two-way ANOVA.